Automating Court Transcripts: Deploying Offline AI Under Strict Government Privacy Laws.
External contractors were drowning in manual audio work. Delays stretched to weeks. I designed and built a fully offline AI transcriber that runs on court hardware, generates same-day dirty transcripts, and opens a direct revenue stream without adding headcount.
& Developer
June 2025
Are you a hiring manager?
Paste your job description and my AI will tell you exactly why I'm a fit.
The Problem vs. The Solution
Government privacy law ruled out every commercial AI tool on the market. So we built one that never touches the cloud. Turnaround dropped from weeks to hours, and the courts walked away with a revenue stream they didn't have before.
The Transcription Backlog
Court clerks were stuck waiting days, sometimes weeks, for external contractors to manually transcribe audio. And every cloud-based AI tool that could have helped? Legally off-limits.
Offline AI Automation
A Python-based desktop app that runs entirely on local hardware, fits into how clerks already work, and spits out a timestamped transcript in hours. No internet required.
The Agile Evolution
Each sprint addressed a distinct constraint. We started by mapping real friction through stakeholder interviews and low-fidelity wireframes, then designed the offline WhisperX + PyAnnote pipeline to satisfy strict data-security requirements, and finally prototyped the Flet desktop interface that transformed raw speaker-diarized audio into a searchable, timestamped transcript output.
Manual Chaos: The Friction.
Courtroom audio was a black box. Over 100 hours of unindexed recordings sat in storage, requiring clerks to manually scrub through tapes. That process took days and cost requesters $9.25 a page.
-
✗
Unsearchable physical media (USB/Tapes)
-
✗
Prohibitive costs for self-represented litigants
-
✗
Inaccessible legalese in procedural filings
Under The Hood
Running production-grade NLP on government hardware with no internet, no GPU, and strict data rules took some creative problem-solving. Here's what's actually happening under the hood.
AI Tech Explainer
Curious how WhisperX and speaker diarization actually work without an internet connection? Pick your background and I'll generate an explanation pitched at exactly your level.
Want to see my UX thinking?
Here's how I broke the transcription problem apart: mapping constraints, failure states, and edge cases, packaged into a custom React slide deck.
View Problem-Solving Logic Close Interactive View
Try the AI Transcriber
This is the real thing. Load a file, watch it process, and explore the full output interface, exactly as it was designed and shipped.
The Business Impact
Going offline-first wasn't a constraint. It was the unlock. Cutting out third-party vendors freed up budget, earned trust from leadership, and gave the courts a new way to generate revenue on their own terms.
Transcription Turnaround Time
Days required to generate a preliminary transcript.
Let's build something impactful.
I'm currently open to new opportunities in UX/UI Design, Product Design, and GovTech transformation roles. If you have a legacy nightmare that needs solving, I'd love to chat.